Authors:
Santiago Salamanca Miño
1
;
Carlos Cerrada Somolinos
2
;
Antonio Adán Oliver
3
and
Miguel Adán Oliver
3
Affiliations:
1
Universidad de Extremadura (UEX), Spain
;
2
Universidad Nacional de Educación a Distancia (UNED), Spain
;
3
Universidad de Castilla la Mancha (UCLM), Spain
Keyword(s):
Matching, recognition, unstructured range data, 3D computer vision.
Related
Ontology
Subjects/Areas/Topics:
Informatics in Control, Automation and Robotics
;
Robotics and Automation
;
Vision, Recognition and Reconstruction
Abstract:
This paper presents a method to estimate the pose (position and orientation) associated to the range data of an object partial view with respect to the complete object reference system. A database storing the principal components of the different partial views of an object, which are generated virtually, is created in advance in order to make a comparison between the values computed in a real view and the stored values. It allows obtaining a first approximation to the searched pose transformation, which will be afterwards refined by applying the Iterative Closest Point (ICP) algorithm. The proposed method obtains very good pose estimations achieving very low failure rate, even in the case of the existence of occlusions. The paper describes the method and demonstrates these conclusions by presenting a set of experimental results obtained with real range data.